System fault probability prediction method and system based on Bayesian network

Abstract

The invention provides a system fault probability prediction method and system based on a Bayesian network, relates to the technical field of aerospace science, and is applied to an SLD light source system. The prediction method comprises the following steps: acquiring environmental parameters and task parameters influencing an SLD light source system, and determining load data borne by the SLD light source system; based on the load data, performing a simulation test on the SLD light source system to obtain performance response data of the SLD light source system under the load data; constructing a Bayesian network model, wherein the Bayesian network model is used for describing a probability relationship between a system state and an external condition; and inputting the target environment parameters and the target task parameters of the to-be-evaluated SLD light source system into the Bayesian network model to obtain fault probability data of the SLD light source system under specified conditions. According to the method, the objectivity and repeatability of the prediction process and the accuracy of the result are improved, and a decision basis is provided for on-orbit life management of a high-reliability system such as an SLD light source.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of aerospace science and technology, in particular, to a system failure probability prediction method and system based on a Bayesian network. BACKGROUND

[0002] The space environment for spacecraft in-orbit operation is very complex, including charged particle radiation, atomic oxygen, micro-meteors and space debris and other factors. Among them, high-energy charged particles (such as electrons, protons, heavy ions) may induce ionization, displacement, single event effects and other radiation effects in the spacecraft electronic system, and may induce charging and discharging effects, which will seriously affect the performance and in-orbit life of the spacecraft. However, different types of systems and electronic devices have different sensitivities to the space environment, and different evaluation systems and methods correspond to different types of radiation effects. The super radiation light emitting diode (SLD, Superluminescent Diode) light source is a kind of semiconductor light source with high power and wide spectrum characteristics, and is widely used in fiber gyroscopes, which is a key component of satellite and spacecraft inertial navigation systems. The SLD itself can be regarded as a subsystem composed of SLD chips, epoxy resin glue, thermistors and refrigerators, and its reliability directly affects the performance of the entire navigation system.

[0003] At present, the research on SLD light source mainly focuses on displacement damage effect, and insufficient attention is paid to the surface and deep charging and discharging effects that may be induced by low-energy electrons in space. Since the SLD light source contains insulating materials such as epoxy resin glue, it has a high risk of charging and discharging in orbit, but the mechanism of charging and discharging effect and its influence on system performance have not been fully evaluated. In addition, if the SLD light source system is simultaneously subjected to the coupling effect of displacement effect and charging and discharging effect, there is currently a lack of effective method for reliability and life evaluation of such devices under different orbital conditions, which is also an important problem in the current technology. SUMMARY

[0004] The present application solves one or more of the above related technical problems.

[0005] To solve the above problems, the present application provides a system failure probability prediction method and system based on a Bayesian network.

[0006] In a first aspect, the present application provides a system failure probability prediction method based on a Bayesian network, applied to an SLD light source system, which comprises: Obtaining one or more groups of environmental parameters and task parameters affecting the SLD light source system; Based on the environmental parameters and the task parameters, determining the corresponding load data borne by the SLD light source system; Based on the load data, a simulation test is performed on the SLD light source system to obtain performance response data of the SLD light source system corresponding to the load data; Based on the environmental parameters, the task parameters, the load data, and the performance response data, a Bayesian network model is constructed, which is used to describe the probability relationship between the system state and the external conditions; The target environmental parameters and the target task parameters of the SLD light source system to be evaluated are input into the Bayesian network model to obtain the failure probability data of the SLD light source system under the specified conditions.

[0007] Optionally, the Bayesian network model includes input nodes representing the environmental parameters and the task parameters, output nodes representing the performance response data, and a conditional probability table storing the probability dependency relationship between the input nodes and the output nodes.

[0008] Optionally, the construction process of the Bayesian network model further includes: The environmental parameters and the task parameters are taken as input variables to establish input nodes of the Bayesian network model; The performance response data is taken as an output variable to establish output nodes of the Bayesian network model; Based on the load data, the probability relationship between the environmental parameters, the task parameters, and the performance response data is determined to establish a conditional probability table connecting the input nodes and the output nodes.

[0009] Optionally, the environmental parameters include spatial environmental parameters, and the spatial environmental parameters include an orbit type and a solar activity state; the task parameters include a task time.

[0010] Optionally, the load data includes cumulative fluence data of different types of particles under a preset confidence level; and the determination of the load data borne by the SLD light source system based on the environmental parameters and the task parameters includes: According to the orbit type, the solar activity state, and the task time, particle flux spectrum data of different types of particles are obtained through a spatial radiation environment analysis model; The particle flux spectrum data are integrated to obtain corresponding average cumulative fluence data; The average cumulative fluence data are corrected according to a preset confidence level to obtain cumulative fluence data of different types of particles under the preset confidence level.

[0011] Optionally, the particles include at least one of high-energy protons, high-energy electrons, and low-energy electrons.

[0012] Optionally, the failure probability data includes the probability of the system experiencing a charge-discharge effect and / or the probability that the system can still function normally after its performance has degraded.

[0013] The beneficial effects of the system failure probability prediction method based on Bayesian networks of the present invention are: By acquiring one or more sets of environmental and task parameters and determining the corresponding load data, simulation experiments are conducted to obtain performance response data. Finally, a Bayesian network model is constructed based on the aforementioned structured data. This method establishes a computable probabilistic mapping relationship from external conditions to system state by parameterizing the actual task scenario, quantifying the physical load, and empirically verifying the performance response. In the application phase, only new target parameters need to be input, and the model can be used to quantitatively infer the failure probability. This process transforms the traditional qualitative extrapolation relying on expert experience and single experimental data into a systematic quantitative evaluation based on multiple sets of structured experimental data and probabilistic graphical models. This significantly improves the objectivity, repeatability, and accuracy of the prediction process, providing a scientific and efficient decision-making basis for the on-orbit lifespan management of high-reliability systems such as SLD light sources.

[0014] To address the above problems, this invention provides a system fault probability prediction system based on Bayesian networks, comprising: The acquisition unit is used to acquire one or more sets of environmental parameters and task parameters that affect the SLD light source system, and to determine the load data borne by the SLD light source system based on the environmental parameters and the task parameters. The experimental unit is used to conduct simulation experiments on the SLD light source system based on the load data, and to obtain the performance response data of the SLD light source system under the load data. The construction unit is used to construct a Bayesian network model based on the environmental parameters, the task parameters, the load data, and the performance response data. The Bayesian network model is used to describe the probabilistic relationship between the system state and external conditions. The processing unit is used to input the target environmental parameters and target task parameters of the SLD light source system to be evaluated into the Bayesian network model to obtain the failure probability data of the SLD light source system under specified conditions.

[0015] The Bayesian network-based system failure probability prediction system and the Bayesian network-based system failure probability prediction method described in this invention have the same advantages over the prior art, and will not be repeated here.

[0016] To address the aforementioned problems, this invention provides a system fault probability prediction device based on Bayesian networks, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the system fault probability prediction method based on Bayesian networks.

[0017] The system fault probability prediction device based on Bayesian networks described in this invention has the same advantages over the prior art as the system fault probability prediction method based on Bayesian networks, and will not be repeated here.

[0018] To address the aforementioned problems, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the system fault probability prediction method based on Bayesian networks.

[0019] The computer-readable storage medium described in this invention has the same advantages over the prior art as the Bayesian network-based system failure probability prediction method, and will not be repeated here. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating a system fault probability prediction method based on Bayesian networks according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the state probability distribution of all nodes in a Bayesian network model during a solar high-age year, according to an embodiment of the present invention. Figure 3 This is a schematic diagram of the state probability distribution of all nodes in a Bayesian network model during a solar low year, according to an embodiment of the present invention. Figure 4 This is a schematic diagram illustrating the relationship between the change in output power of an SLD light source and the irradiation dose under 5MeV proton-electron irradiation conditions, according to an embodiment of the present invention. Figure 5 This is a schematic diagram illustrating the relationship between the change in output power of an SLD light source and the irradiation dose under 1MeV electron irradiation conditions, according to an embodiment of the present invention. Figure 6 This is a schematic diagram illustrating the relationship between the change in output power of an SLD light source and the irradiation dose under 110keV electron irradiation, according to an embodiment of the present invention. Figure 7 This is a schematic diagram of the equivalent flux of an SLD light source system under different aluminum shielding layer thicknesses in a 10-year GEO orbit according to an embodiment of the present invention. Figure 8 This is a schematic diagram of the equivalent flux of an SLD light source system under different aluminum shielding layer thicknesses in a 20-year GEO orbit according to an embodiment of the present invention. Figure 9 This is a schematic diagram of the high-energy electron flux accumulated in the GEO orbit over 10 hours under different aluminum shielding layer thicknesses in an SLD light source system according to an embodiment of the present invention. Detailed Implementation

[0021] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Although some embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present invention. It should be understood that the accompanying drawings and embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0022] It should be understood that the various steps described in the method embodiments of the present invention may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this respect.

[0023] The term "comprising" and its variations as used herein are open-ended, meaning "including but not limited to"; the term "based on" means "at least partially based on"; the term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments"; and the term "optionally" means "optional embodiments". Definitions of other terms will be given in the following description. It should be noted that the concepts of "first," "second," etc., mentioned in this invention are only used to distinguish different devices, functional modules, or units, and are not intended to limit the order of functions performed by these devices, functional modules, or units or their interdependencies.

[0024] It should be noted that the terms "a" and "a plurality of" used in this invention are illustrative rather than restrictive. Those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0025] The names of the messages or information exchanged between the multiple devices in the embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of these messages or information.

[0026] In high-reliability fields such as aerospace and energy, system failures are often caused by the combined effects of multiple environmental factors (such as radiation, temperature, and particle bombardment) and mission profiles (such as operating time and orbital conditions), with complex coupling and uncertainties among these factors. Traditional reliability assessment methods are usually based on a single failure mechanism or deterministic model, making it difficult to comprehensively characterize the system degradation process under multi-stress coupling, nor can they effectively quantify the impact of environmental uncertainties and system response randomness on failure probability. For example, when assessing the impact of the space radiation environment on the reliability of a certain type of light source system, existing methods often analyze different effects such as displacement damage and charging / discharging independently, or extrapolate based solely on single ground test data. There is a lack of an assessment framework that can integrate multi-source environmental data and test response data, and systematically process their inherent probabilistic correlations. This leads to prediction results that often deviate significantly from actual on-orbit conditions, making it difficult to support accurate lifetime prediction and reliability design.

[0027] Therefore, there is an urgent need for a method that can integrate environmental parameters, mission conditions, and experimental data, and use probabilistic models to handle uncertainties, ultimately achieving quantitative failure probability prediction.

[0028] To address the problems existing in the aforementioned related technologies, embodiments of the present invention provide a system and method for predicting system failure probability based on Bayesian networks.

[0029] like Figure 1 As shown in the figure, an embodiment of the present invention provides a system fault probability prediction method based on Bayesian networks, applied to an SLD light source system. The system fault probability prediction method based on Bayesian networks includes: Step S100: Obtain one or more sets of environmental parameters and task parameters that affect the SLD light source system.

[0030] Specifically, by systematically collecting various environmental and mission parameters related to the on-orbit operation of the SLD light source system, the acquired parameters are not limited to single-point values ​​but can cover combinations of various typical or boundary conditions. For example, environmental parameters mainly include variables describing the natural physical conditions of space in which the SLD light source system is located, such as orbit type (e.g., GEO / MEO / LEO), solar activity status (e.g., high-year / low-year), space radiation background, atomic oxygen flux, etc. These parameters directly affect the type and intensity of physical loads the system endures. Mission parameters mainly include engineering variables describing the mission design of the spacecraft or payload itself, such as mission duration (on-orbit lifetime), operating mode (continuous / intermittent), attitude pointing, thermal control conditions, etc. These parameters determine the duration and manner in which the system is exposed to the aforementioned environmental conditions. These parameters together constitute a set of external variables affecting system reliability, and their diversity ensures that subsequent models can learn and characterize system behavior under a wider range of scenarios.

[0031] By actively collecting multiple sets of environmental and mission parameters, a structured, multi-scenario input condition library is provided for the entire prediction method, effectively avoiding evaluation bias caused by single or insufficient data coverage. This not only enables subsequent payload calculations and experimental design to be based on more representative external conditions, but also lays a crucial data foundation for building a probabilistic prediction model with good generalization capabilities, thereby significantly improving the adaptability of the method to complex and variable on-orbit environments and the reliability of evaluation results.

[0032] Step S200: Based on the environmental parameters and the task parameters, determine the corresponding load data that the SLD light source system bears.

[0033] Specifically, the acquired environmental and mission parameters are used as inputs and, through a pre-defined physical model or data analysis method, are transformed into quantified load data that the system experiences during actual operation. This process is essentially a mapping from "external condition description" to "internal effect quantification." For example, based on orbital parameters, time parameters, and a space environment model, physical quantities such as the cumulative radiation fluence expected to be experienced by the system during a specific mission can be calculated. The output of this step is load data characterizing the specific physical impact of external conditions on the system, which can be directly used for subsequent experiments or simulations.

[0034] The core function of this step is to establish a crucial bridge between multidimensional, macroscopic environmental and task descriptions and specific, quantifiable, and reproducible physical loads. By transforming complex integrated external conditions into explicit engineering load data, it not only provides precise input for subsequent simulation experiments, ensuring the authenticity and relevance of experimental conditions, but also makes the effects under different environmental conditions comparable and calculable. This fundamentally solves the problem of reliability assessment bias caused by unclear loads and lays a solid physical foundation for building accurate probabilistic prediction models.

[0035] Step S300: Based on the load data, conduct a simulation test on the SLD light source system to obtain the performance response data of the SLD light source system under the load data.

[0036] Specifically, based on defined load data, controlled simulation experiments are conducted on the SLD light source system or a representative sample thereof. This experiment aims to reproduce or equivalently represent the physical loads (such as radiation) experienced by the system in actual tasks, and to monitor and record the changes in key performance indicators (such as output optical power) of the system as the load applies. In this way, quantitative response data of system performance as a function of load are obtained, establishing a true physical correlation between "load-performance response".

[0037] Through controlled simulation experiments, an empirical correlation was directly established between quantified loads and actual system performance degradation, transforming theoretically or model-calculated loads into real, observable system response data. This effectively compensates for the shortcomings of pure theoretical analysis or environmental data extrapolation, providing crucial and reliable training and validation data for subsequent probabilistic prediction model construction, and significantly enhancing the empirical foundation and credibility of the entire prediction method.

[0038] Step S400: Based on the environmental parameters, the task parameters, the load data, and the performance response data, a Bayesian network model is constructed. The Bayesian network model is used to describe the probabilistic relationship between the system state and external conditions.

[0039] Specifically, based on the structured data acquired in steps S100 to S300, a Bayesian network probabilistic graphical model is constructed. This model uses quantified environmental and task conditions, physical loads, and system performance responses as variables, forming a network structure by defining causal or conditional dependencies between these variables. Subsequently, experimental data and prior knowledge are used to probabilistically characterize the strength of these dependencies (e.g., the probability of system performance degradation to some extent under specific radiation load data). This model integrates multi-source heterogeneous data within a unified probabilistic reasoning framework, and its final form is a quantitative tool capable of taking specific environmental and task conditions as input and outputting the probability distribution of the system in various states (e.g., normal, fault, performance degradation).

[0040] By constructing a Bayesian network model, this approach organically combines physical mechanisms with statistical inference, enabling a quantitative and structured characterization of multi-factor, nonlinear, and uncertain system failure processes. Its core value lies in its ability to systematically integrate prior knowledge, environmental data, and experimental results, and to utilize probabilistic reasoning to handle scenarios with incomplete information or noise. This allows the model to not only output point-estimated failure probabilities but also provide a complete probability distribution, thereby supporting risk assessment and decision-making. The model exhibits good interpretability and updability; when new data is obtained, the prediction results can be dynamically updated, significantly improving the adaptability, robustness, and scientific decision support capabilities of failure prediction.

[0041] Step S500: Input the target environmental parameters and target task parameters of the SLD light source system to be evaluated into the Bayesian network model to obtain the failure probability data of the SLD light source system under specified conditions.

[0042] Specifically, when a reliability assessment is required for a new SLD (Solar Lamp Discharge) light source mission scenario (e.g., a satellite planned for launch into a new orbit), the target environmental parameters (such as orbit type and expected solar activity state) and target mission parameters (such as design life) are input into a pre-built Bayesian network model. After receiving these inputs, the model automatically performs probability propagation and inference using its internally learned probabilistic relationship network extracted from historical data, ultimately outputting the probability value of the system failing in that target scenario (such as excessive degradation of output optical power or charging / discharging). This process achieves a direct mapping from "scenario description" to "risk quantification."

[0043] By directly inputting mission design parameters, failure probabilities can be obtained in real time, eliminating the enormous cost and time of repeatedly performing complex environmental analyses, load calculations, and experimental designs for each new scenario. This greatly improves the efficiency and agility of reliability assessment, making rapid risk iteration and solution optimization possible in the early stages of mission design, and providing powerful data-driven decision support for spacecraft system design and mission planning.

[0044] In this embodiment, one or more sets of environmental and task parameters are acquired, and corresponding load data is determined based on them. Simulation experiments are then conducted to obtain performance response data, and finally, a Bayesian network model is constructed based on the aforementioned structured data. This method establishes a computable probabilistic mapping relationship from external conditions to system state by parameterizing the actual task scenario, quantifying the physical load, and empirically verifying the performance response. In the application phase, only new target parameters need to be input, and the model can be used to quantitatively infer the failure probability. This process transforms the traditional qualitative extrapolation relying on expert experience and single experimental data into a systematic quantitative evaluation based on multiple sets of structured experimental data and probabilistic graphical models. This significantly improves the objectivity, repeatability, and accuracy of the prediction process, providing a scientific and efficient decision-making basis for the on-orbit lifespan management of high-reliability systems such as SLD light sources.

[0045] Optionally, the Bayesian network model includes input nodes representing environmental parameters and task parameters, output nodes representing system performance response data, and a conditional probability table storing the probabilistic dependencies between the input nodes and the output nodes.

[0046] Optionally, the process of constructing the Bayesian network model further includes: Using the environmental parameters and the task parameters as input variables, the input nodes of the Bayesian network model are established; Using the performance response data as the output variable, the output node of the Bayesian network model is established; Based on the load data, the probabilistic correlation between the environmental parameters, the task parameters, and the performance response data is determined to establish a conditional probability table connecting the input node and the output node.

[0047] Specifically, first, the data foundation and objective of the model are clarified, namely, using all related data to characterize the complete probabilistic chain from external conditions to system state. Second, the core structure of the model is defined. This network model consists of three types of nodes: input nodes representing causes or conditions (such as environmental and task parameters), output nodes representing results or states (such as performance response), and conditional probability tables storing the quantified influence relationships between nodes. The specific construction steps are as follows: mapping environmental and task parameters to input nodes, mapping performance response data to output nodes, and filling and determining the conditional probability tables connecting these nodes based on the observed statistical or physical relationships between load data and performance response data, thereby completing a network model capable of probabilistic inference. Load data is not directly used as an input node. It serves as an intermediate computational quantity or feature variable connecting environmental / task parameters (input nodes) and performance response (output nodes). In constructing the conditional probability tables, the role of load data is to act as a medium for quantifying the intensity or mechanism of influence.

[0048] Input nodes are the direct cause variables of the model, typically observable or programmable conditions such as orbit type, solar activity status, mission duration, and shielding thickness. These are parameters that engineers can directly specify or obtain from mission design.

[0049] Payload data (such as cumulative proton fluence and low-energy electron fluence) are intermediate physical quantities calculated from input nodes (environmental / mission parameters) using a physical model. They represent the specific physical effects of external conditions on the system.

[0050] The most crucial step is establishing the conditional probability table. This involves not simply inputting the load data as a node, but rather analyzing the statistical relationship between the load data and the performance response data, and then attributing or mapping this relationship to the combination of input nodes that generated the load.

[0051] The specific technical logic is as follows: Relationship chain: Input node combination → (calculated through physical model) → specific load data → (observed through experiments) → specific performance response distribution.

[0052] Reverse learning during modeling: When building a model, we have multiple sets of data: {input node values, calculated load values, observed performance response values}.

[0053] Establishing CPT: We use load data as a bridge to analyze: "When the input node is combination A, the calculated load is value L; while in the experiment, when the load is value L, we observe that the probability of the performance response being state R is P." Ultimately, what we record in the conditional probability table (CPT) is: when the input node is combination A, the probability that the output node is state R is P.

[0054] Here, the payload data L is the key calculation and experimental basis for helping us determine the probability P, but it does not exist as an explicit node in the network.

[0055] For example: Suppose the input node is: solar activity state = {high, low}, shielding thickness = {thick, thin}.

[0056] The output node is: Optical power state = {Good, Degraded}.

[0057] Step 1 (Physical Calculation): When solar activity is high and shielding thickness is thin, the proton fluence is calculated to be high dose using a radiation model.

[0058] Step 2 (Experimental Observation): In the laboratory, the samples were irradiated with a high dose of protons. It was observed that 80% of the samples had a degraded optical power state and 20% had a good optical power state.

[0059] Step 3 (Establishing CPT): In the conditional probability table of the Bayesian network, we directly record: When solar activity is high and shielding thickness is thin, the probability of optical power degradation is 80%, and the probability of good optical power is 20%.

[0060] The calculated proton fluence of high dose is the basis for the 80% probability value, but it does not appear as a node in the final inference network.

[0061] The above process describes the offline construction and training phase of a Bayesian network model. Its core lies in utilizing a systematically prepared set of prior data to transform the causal relationships in the physical world into a computable probabilistic graphical model. Specifically: Data foundation: Integrating four key data sets: environmental parameters (such as various orbital conditions and solar activity status), mission parameters (such as different mission durations), payload data (i.e., physical quantities such as radiative flux calculated based on the former two), and performance response data (i.e., performance degradation data exhibited by the system under different payload data in ground simulation experiments).

[0062] Structure definition: Based on domain knowledge, determine the network topology, that is, which factors are "causes" (as input nodes, such as environment and task parameters), which states are "effects" (as output nodes, such as system performance level), and clarify the dependencies between them.

[0063] Parameter learning: The core step is to quantify the uncertainty of the statistical or physical correspondence between load data and performance response data through analysis and calculation, thereby filling in a conditional probability table. This table precisely expresses "the probability that the output will be a certain state under given input conditions".

[0064] For example, to construct a Bayesian network model for evaluating the on-orbit reliability of SLD light sources, a database first needs to be established. Environmental and mission parameters may include various combinations such as {Geostationary Orbit (GEO), high solar activity year, 5-year mission period} and {Medium Earth Orbit (MEO), low solar activity year, 10-year mission period}. For each combination, the corresponding cumulative proton and electron fluence is calculated using a radiation environment model to form payload data. Subsequently, in the laboratory, irradiation tests are conducted on a batch of representative SLD light source samples according to these fluence levels, and the attenuation data of the output optical power of each sample is recorded to form performance response data. When constructing the model, the algorithm analyzes statistical patterns such as "when the input is GEO + high solar activity year + 5 years, the corresponding fluence level in the payload data is B; and when the fluence is B, the proportion of samples showing an optical power decrease of more than 20% in the performance response data is X%", and fills X% as a conditional probability into the table. Subsequently, to evaluate a brand-new SLD light source planned to operate in GEO orbit for 15 years, engineers only need to input the designed parameters into this trained model, and the model can instantly output a probability prediction of a severe drop in its optical power based on the internalized probability relationship, without having to wait for 15 years of real on-orbit data or conduct a complete long-term ground test.

[0065] like Figure 2 and Figure 3 As shown, the specific output results and empirical evidence for the SLD light source system in a typical application scenario (GEO track) are presented.

[0066] in, Figure 2 This is a schematic diagram of the state probability distribution of all nodes in a Bayesian network model during the solar high years. Figure 3This diagram illustrates the state probability distribution of all nodes in a Bayesian network model during a low solar activity year. In the diagram, the "Solar_activity" node represents the solar activity state, "Mission_time" represents the on-orbit operating time, "Shielding" represents the aluminum shielding layer thickness, "Dd" represents the equivalent flux under given conditions, "Charge" indicates the presence of a charging / discharging effect, and "Power_out" represents the output optical power of the SLD light source. When the aluminum shielding layer thickness is "little" (0.01 mm), the probability of a charging / discharging effect is 100%. When the shielding layer thickness is "high" (2 mm), there is essentially no charging / discharging effect, and displacement damage is the primary concern. During a high solar activity year, the probability of the system operating normally is 73.7%, while during a low solar activity year, this value is 89.5%, demonstrating the significant impact of solar activity on the system. As the shielding layer thickness continues to increase, the degree of displacement damage decreases accordingly, and the probability of a charging / discharging effect drops to 0.

[0067] Figure 2 and Figure 3 The Bayesian network model in this model is specifically designed to evaluate the on-orbit reliability of SLD light source systems. The node Power_out (output optical power) directly corresponds to the core performance indicator of the SLD light source system. The node Charge (charge and discharge effect) is designed to address the unique failure risks associated with the insulating materials such as epoxy resin within the SLD light source. It provides quantified failure probability predictions. For example, under high solar conditions, the probability of the SLD light source system's output power being in a "Good" state is 73.7%, the probability of it being "Degraded" is 15.6%, and the probability of it being "Bad" is 10.7%. This directly transforms abstract risks into concrete probabilistic data that can support engineering decisions.

[0068] By comparing the graphs from high solar years and low solar years, the significant impact of solar activity on system reliability can be clearly seen. During low solar years, the probability of good output power increases to 89.5%. This intuitively demonstrates how Bayesian networks can quantify the impact of different environmental factors (solar activity) on the final failure probability of the SLD light source system.

[0069] In summary, Figure 2 and Figure 3 This document presents the specific output results and empirical evidence of the Bayesian network-based system failure probability prediction method for SLD light source systems in a typical application scenario (GEO track). It demonstrates how environmental and design parameters can be transformed into quantitative predictions of the operating status (good, degraded, or faulty) of the SLD light source system through a probabilistic model, perfectly reflecting the technical effectiveness and practical value of this invention.

[0070] The core advantage of the Bayesian network model construction method lies in its fusion modeling of physical failure mechanisms and statistical correlations, unifying multi-source heterogeneous data into a computable probabilistic graphical framework. By explicitly distinguishing environmental and task-related "causal" variables from system performance "effect" variables as network nodes and quantifying their influence relationships using conditional probability tables, the model is not only structurally clear and highly interpretable but also naturally handles multi-factor coupling effects and uncertainty propagation. It allows for the probabilistic connection of limited data obtained from ground tests (load-response relationships) with broad and diverse on-orbit environmental conditions (input node states), thereby enabling the extrapolation and prediction of system failure probabilities under various on-orbit scenarios using limited experimental data. This significantly enhances the flexibility, adaptability, and engineering practical value of the evaluation method.

[0071] Optionally, the environmental parameters include space environment parameters, which include orbit type, solar activity status, and shielding layer thickness; the mission parameters include mission time.

[0072] Optionally, the load data includes cumulative flux data of different types of particles under a preset confidence level; determining the load data borne by the SLD light source system based on the environmental parameters and the task parameters includes: Based on the orbit type, the solar activity status, and the mission duration, particle flux spectrum data for different types of particles are obtained using a space radiation environment analysis model. The particle flux spectrum data is integrated to obtain the corresponding average cumulative flux data; The average cumulative injection data of each type of particle are corrected according to the preset confidence level to obtain the cumulative injection data of different types of particles under the preset confidence level.

[0073] Optionally, the particle type includes at least one of high-energy protons, high-energy electrons, and low-energy electrons.

[0074] Specifically, firstly, key input parameters were defined: space environment parameters (such as orbit type, solar activity state, and shielding thickness) and mission parameters (such as mission duration). These parameters were input into a specialized space radiation environment analysis model (such as SPENVIS) to calculate particle flux data for a specific energy spectrum. By integrating this flux data over energy and time dimensions, the average cumulative flux during the mission period was obtained. To reflect the inherent uncertainties of the space environment, this average flux was statistically corrected based on commonly used engineering confidence levels (such as 90% and 95%), ultimately yielding cumulative flux data containing uncertainty information for subsequent experiments and evaluations. The payload was specifically characterized by the flux of different types of particles, such as high-energy protons, high-energy electrons, and low-energy electrons.

[0075] For example, to evaluate a scenario where an SLD (Solar Light Source) is in geosynchronous orbit (GEO), experiencing a high solar activity year, using 2mm aluminum shielding, and performing a 15-year mission, the following steps are taken: First, input these parameters (orbit type = GEO, solar activity state = high solar activity year, shielding thickness = 2mm, mission duration = 15 years) into a space environment model. The software will output the proton and electron flux spectra received per unit area per unit time in that orbit at different energies. Next, integrate this flux spectrum over the 15-year mission period and within the energy range of interest to obtain the average total fluence expected to be received during the mission period. Finally, based on engineering requirements (e.g., using a 95% confidence level), and through statistical uncertainty analysis of the space environment model, determine the appropriate confidence factor to correct the average fluence, obtaining the cumulative fluence data at a 95% confidence level for guiding the design of ground-based irradiation experiments, for example, 5 MeV protons with a fluence of not less than 1 × 10⁻⁶. 11 p / cm 2 This data will serve as a direct basis for the irradiation dose in subsequent simulation experiments. For example... Figure 4 and Figure 5 The figures show schematic diagrams illustrating the relationship between the output power of the SLD source and the irradiation flux under 5MeV proton and 1MeV electron irradiation conditions, respectively. Figure 4 and Figure 5 It can be seen that as the irradiation dose increases, the output optical power gradually decreases. For example... Figure 6 This diagram illustrates the relationship between the output power of an SLD light source and the irradiation flux under 110 keV electron irradiation. Figure 6 It can be seen that the system exhibits a discharge effect when the injection volume reaches approximately 10⁹.

[0076] Among them, high-energy protons, high-energy electrons and low-energy electrons were selected for irradiation experiments. The energy and flux parameters of high-energy protons, high-energy electrons and low-energy electrons were determined by the results of orbital environment analysis, that is, the cumulative flux calculated according to the mission orbit and lifetime in the corresponding energy range, and selected in combination with the representative energy that the test device can provide and the necessary engineering margin.

[0077] Subsequently, experiments were conducted based on the maximum particle flux received under different orbital conditions. The output optical power was measured multiple times during the experiments to determine the relationship between optical power variation and irradiation flux. The flux uniformity error within the irradiated area of ​​the tested sample should be less than 10%. High-energy particle flux has a relatively small impact on sample displacement damage and can be as low as 10⁸ particles / cm². 2 ·s ~ 10⁹ particles / cm 2 The choice should be between ·s and ·s, with the low-energy electron flux controlled at 10⁹ particles / cm². 2 Below ·s, to prevent large instantaneous charge accumulation from causing discharge.

[0078] In some embodiments, the process is detailed as follows: starting from the spacecraft's mission design parameters, a standardized space environment engineering analysis workflow is used to calculate the key radiation load data (i.e., particle cumulative fluence) for reliability assessment. Specifically: Input parameter determination: First, the basic mission design parameters of the spacecraft housing the SLD light source system are defined, including orbital parameters (such as orbital altitude and inclination), mission lifetime, and space environment conditions that require special consideration (such as distinguishing between high and low solar activity years, and determining whether the orbit crosses high-radiation regions such as the South Atlantic Anomaly). These parameters form the basis for all subsequent analyses.

[0079] Environmental model calculation: Input the above parameters into professional space radiation environment analysis software (such as SPENVIS or OMERE). In the software, select an industry-recognized standard environmental model based on the analysis object (proton or electron) (e.g., AP-8 model for protons, AE-8 model for electrons). After the software runs, it outputs a key intermediate result: the particle flux spectrum Φ(E,t). This spectrum characterizes the number of incident particles per unit time and unit energy interval at the orbital position during the mission, and is a function of energy (E) and time (t).

[0080] Integrating to obtain cumulative action: Double integral over the flux spectrum to convert it into total action: Energy integral: within the energy range of interest (from the minimum energy E) min To the maximum energy E max Within a certain time, the flux spectrum is integrated to obtain the total flux of the energy spectrum at a certain moment.

[0081] Time integration: Integrate the above results over the entire mission lifetime T.

[0082] The final integral formula is expressed as: ; The calculated result N is the average cumulative flux expected to be received during the mission, typically expressed in particles per square centimeter (particles / cm²). 2 ).

[0083] Confidence Level Correction: Due to the inherent uncertainties in the space environment model, statistical correction is required to the average flux to meet the conservative principles necessary for engineering reliability design. Based on relevant standards or engineering experience, an appropriate confidence level coefficient (e.g., corresponding to 90% or 95% confidence levels) is selected. The calculated average cumulative flux N is then multiplied by this coefficient to obtain the design baseline flux used to guide ground testing and lifetime prediction, such as the "cumulative flux at 95% confidence level".

[0084] The process employed recognized software tools (SPENVIS / OMERE) and radiation models (AP-8 / AE-8) in the aerospace field, ensuring the authority and reproducibility of the analysis results.

[0085] The physical meaning is clear: through double integration, dynamic and energy-spectralized environmental data is condensed into a single, total radiation load index (cumulative flux), which is directly related to subsequent radiation effects (such as displacement damage).

[0086] Engineering applicability: By introducing a confidence level correction step, the uncertainty of the scientific model is transformed into a safety margin in engineering, so that the final output load data can directly serve the high reliability design verification and life assessment.

[0087] This calculation step is the hub connecting the "mission scenario" and the "physical experiment" in the whole method. It ensures that the radiation dose used in the ground simulation experiment can realistically and conservatively reflect one of the most severe radiation environments that may be encountered during the on-orbit mission, thereby ensuring the effectiveness of all subsequent reliable predictions based on experimental data.

[0088] The significant benefit of this process lies in achieving a standardized and quantifiable mapping from macroscopic mission scenarios to microscopic physical effects, and incorporating an uncertainty management mechanism. By rigorously transforming engineering concepts such as "orbit" and "solar activity" into a measurable and reproducible physical quantity called "proton / electron fluence" using recognized environmental models, an objective and unified payload benchmark is provided for the entire evaluation process. Simultaneously, the introduction of a confidence level correction step scientifically covers the inherent biases and fluctuations of space environment models, ensuring that the output payload data is no longer a single, definitive value, but rather a boundary value with statistical guarantees (such as the maximum expected fluence at a 95% confidence level). This directly supports the subsequent reliability verification approach of "using the worst-case scenario as a benchmark," greatly enhancing the conservatism and engineering credibility of the lifetime prediction results.

[0089] In some embodiments, such as Figure 7 and Figure 8 The diagram shows the equivalent flux of the SLD light source system under different aluminum shielding thicknesses in 10-year and 20-year GEO orbits. Solid curves represent years of high solar activity, and hollow curves represent years of low solar activity. Each curve in the diagram corresponds to a cumulative distribution function of the equivalent flux under a shielding thickness. Figure 9 This diagram illustrates the accumulated high-energy electron fluence of the GEO orbit over 10 hours under different aluminum shielding thicknesses. (Diagram of the accumulated high-energy electron fluence of the GEO orbit over 10 hours in an SLD light source system under different aluminum shielding thicknesses). The radiation dose D is shown below. d Equivalent radiation dose (RDD) is a key physical quantity that measures the energy deposited by radiation in a material and the resulting damage. In this application, it specifically refers to the energy deposited in the internal material of the SLD light source system by high-energy charged particles (such as protons and electrons) after penetrating the spacecraft's shielding layer; the unit is rad.

[0090] Optionally, the failure probability data includes the probability of the system experiencing a charge-discharge effect and / or the probability that the system can still function normally after its performance has degraded.

[0091] Specifically, the failure probability data includes two core risk assessment dimensions. The first is the probability of a specific physical failure mode occurring, i.e., the probability of the system experiencing a charge-discharge effect. This probability directly stems from the unique risks of insulating materials in SLD light source systems under low-energy electronic environments. The second is the probability of the system maintaining functional integrity after experiencing performance degradation, i.e., the probability that the system can still meet minimum operating requirements ("normal operation") after performance degradation caused by displacement damage, etc. These two types of probability data are not independent; they together constitute an integrated description of the system state: the former focuses on "whether a sudden failure event will occur," while the latter assesses "the system's functional margin under progressive damage."

[0092] By clearly distinguishing between the probability of failure and the probability of survival after damage, multi-dimensional and refined quantitative assessment of system risk is achieved. This directly meets the urgent need in engineering practice for differentiated management of different failure modes. For example, the probability of charge-discharge effects helps guide targeted insulation design or protection strategies; while the probability of normal operation after performance degradation provides a direct basis for judging system life margin and setting performance degradation management thresholds. This distinction makes the prediction results no longer a general judgment of good or bad, but a precise and actionable decision input to guide reliability design, on-orbit health management, and maintenance strategy formulation, greatly improving the practicality and engineering guidance value of reliability assessment results.

[0093] This invention provides a system fault probability prediction system based on Bayesian networks, comprising: The acquisition unit is used to acquire one or more sets of environmental parameters and task parameters that affect the SLD light source system, and to determine the load data borne by the SLD light source system based on the environmental parameters and the task parameters. The experimental unit is used to conduct simulation experiments on the SLD light source system based on the load data, and to obtain the performance response data of the SLD light source system under the load data. The construction unit is used to construct a Bayesian network model based on the environmental parameters, the task parameters, the load data, and the performance response data. The Bayesian network model is used to describe the probabilistic relationship between the system state and external conditions. The processing unit is used to input the target environmental parameters and target task parameters of the SLD light source system to be evaluated into the Bayesian network model to obtain the failure probability data of the SLD light source system under specified conditions.

[0094] This invention provides a system fault probability prediction device based on Bayesian networks, comprising a memory and a processor; the memory is used to store a computer program; the processor is used to implement the system fault probability prediction method based on Bayesian networks as described above when the computer program is executed.

[0095] This invention provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the system fault probability prediction method based on Bayesian networks as described above.

[0096] While the present invention has been disclosed above, its scope of protection is not limited thereto. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention, and all such changes and modifications will fall within the scope of protection of the present invention.

Claims

1. A method for predicting the probability of system failures based on Bayesian networks, characterized in that, The Bayesian network-based system failure probability prediction method, applied to SLD light source systems, includes: Obtain one or more sets of environmental and task parameters that affect the SLD light source system; Based on the environmental parameters and the task parameters, the corresponding load data that the SLD light source system bears is determined; Based on the load data, a simulation experiment was conducted on the SLD light source system to obtain the performance response data of the SLD light source system under the load data. Based on the environmental parameters, the task parameters, the load data, and the performance response data, a Bayesian network model is constructed. The Bayesian network model is used to describe the probabilistic relationship between the system state and external conditions. The target environmental parameters and target task parameters of the SLD light source system to be evaluated are input into the Bayesian network model to obtain the failure probability data of the SLD light source system under specified conditions.

2. The system fault probability prediction method based on Bayesian networks according to claim 1, characterized in that, The Bayesian network model includes input nodes representing environmental and task parameters, output nodes representing system performance response data, and a conditional probability table storing the probabilistic dependencies between the input nodes and the output nodes.

3. The system fault probability prediction method based on Bayesian networks according to claim 2, characterized in that, The process of constructing the Bayesian network model also includes: Using the environmental parameters and the task parameters as input variables, the input nodes of the Bayesian network model are established; Using the performance response data as the output variable, the output node of the Bayesian network model is established; Based on the load data, the probabilistic correlation between the environmental parameters, the task parameters, and the performance response data is determined to establish a conditional probability table connecting the input node and the output node.

4. The system fault probability prediction method based on Bayesian networks according to claim 1, characterized in that, The environmental parameters include space environment parameters, which include orbit type and solar activity status; the mission parameters include mission time.

5. The system fault probability prediction method based on Bayesian networks according to claim 4, characterized in that, The load data includes cumulative flux data of different types of particles under a preset confidence level; determining the load data borne by the SLD light source system based on the environmental parameters and the task parameters includes: Based on the orbit type, the solar activity status, and the mission duration, particle flux spectrum data for different types of particles are obtained using a space radiation environment analysis model. The particle flux spectrum data is integrated to obtain the corresponding average cumulative flux data; The average cumulative injection data of each type of particle are corrected according to the preset confidence level to obtain the cumulative injection data of different types of particles under the preset confidence level.

6. The system fault probability prediction method based on Bayesian networks according to claim 5, characterized in that, The particles include at least one of high-energy protons, high-energy electrons, and low-energy electrons.

7. The system fault probability prediction method based on Bayesian networks according to claim 1, characterized in that, The failure probability data includes the probability of the system experiencing a charge-discharge effect and / or the probability that the system can still function normally after its performance has degraded.

8. A system failure probability prediction system based on Bayesian networks, characterized in that, The Bayesian network-based system failure probability prediction system, applied to SLD light source systems, includes: The acquisition unit is used to acquire one or more sets of environmental parameters and task parameters that affect the SLD light source system, and to determine the load data borne by the SLD light source system based on the environmental parameters and the task parameters. The experimental unit is used to conduct simulation experiments on the SLD light source system based on the load data, and to obtain the performance response data of the SLD light source system under the load data. The construction unit is used to construct a Bayesian network model based on the environmental parameters, the task parameters, the load data, and the performance response data. The Bayesian network model is used to describe the probabilistic relationship between the system state and external conditions. The processing unit is used to input the target environmental parameters and target task parameters of the SLD light source system to be evaluated into the Bayesian network model to obtain the failure probability data of the SLD light source system under specified conditions.

9. A system fault probability prediction device based on Bayesian networks, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the system failure probability prediction method based on Bayesian networks as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the system fault probability prediction method based on Bayesian networks as described in any one of claims 1 to 7.

Images